I love AI cuz it makes millions dollars for me <3
what do you think about ai?
posted in
Off Topic
hYm3nSl4yrdelirentsIMHO, the future is local models and cheaper hardware, so everyone can run AI themselves. But that would mean $0 for their magical cloud compute which is the only business they have.
i'm not sure that hardware can improve at a rate faster than the models grow. it really sucks to have to send your data over to the big companies so they can run inference (and do whatever they do with all the data)
and then there's the problem of efficiency/environmental impact, which people are making a big deal about. But running the models in datacentres is always going to be more efficient / better for the environment than people running the models locally.
really sucks
some analysts predict that local AI models can be as good as AI that ran in a datacenter half a year ago, it's not just the hardware improving but several other factors too like compression techniques and algorithmic efficiencies
a GTX 3090 can run qwen 3.8-27b which gets a better AI intelligence score (doesn't just include googling information) than Claude Opus 4.5, GPT-5.2 and Gemini 3 7 months ago
[quote=hYm3nSl4yr][quote=deli][quote=rents]
IMHO, the future is local models and cheaper hardware, so everyone can run AI themselves. But that would mean $0 for their magical cloud compute which is the only business they have.[/quote][/quote]
i'm not sure that hardware can improve at a rate faster than the models grow. it really sucks to have to send your data over to the big companies so they can run inference (and do whatever they do with all the data)
and then there's the problem of efficiency/environmental impact, which people are making a big deal about. But running the models in datacentres is always going to be more efficient / better for the environment than people running the models locally.
really sucks[/quote]
some analysts predict that local AI models can be as good as AI that ran in a datacenter half a year ago, it's not just the hardware improving but several other factors too like compression techniques and algorithmic efficiencies
a GTX 3090 can run qwen 3.8-27b which gets a better AI intelligence score (doesn't just include googling information) than Claude Opus 4.5, GPT-5.2 and Gemini 3 7 months ago
IMHO, the future is local models and cheaper hardware, so everyone can run AI themselves. But that would mean $0 for their magical cloud compute which is the only business they have.[/quote][/quote]
i'm not sure that hardware can improve at a rate faster than the models grow. it really sucks to have to send your data over to the big companies so they can run inference (and do whatever they do with all the data)
and then there's the problem of efficiency/environmental impact, which people are making a big deal about. But running the models in datacentres is always going to be more efficient / better for the environment than people running the models locally.
really sucks[/quote]
some analysts predict that local AI models can be as good as AI that ran in a datacenter half a year ago, it's not just the hardware improving but several other factors too like compression techniques and algorithmic efficiencies
a GTX 3090 can run qwen 3.8-27b which gets a better AI intelligence score (doesn't just include googling information) than Claude Opus 4.5, GPT-5.2 and Gemini 3 7 months ago
chat gpt told me the owner of scrap.tf was a guy named "scrappy"
chat gpt told me the owner of scrap.tf was a guy named "scrappy"
Literally 3 days before openai announced the navier-stokes stuff, (because there were rumors about it) I was commenting to someone "if someone actually uses ai to prove navier-stokes, I will consider LLMs one of the more substantial technological developments ever."
(granted they found a counter-example, which is perhaps less impressive, since counter-examples seem to be the more common success story for LLMs looking at big problems, since they are presumably easier to "brute-force." + some of the drama w.r.t. the external math researchers uploading their work into chatgpt, and it possibly being used to train the new model...)
quintoshsome analysts predict that local AI models can be as good as AI that ran in a datacenter half a year ago, it's not just the hardware improving but several other factors too like compression techniques and algorithmic efficiencies. A GTX 3090 can run qwen 3.8-27b which gets a better AI intelligence score (doesn't just include googling information) than Claude Opus 4.5, GPT-5.2 and Gemini 3 7 months ago
IIRC, qwen has to use a ton of tokens to get that quality on the benchmark. So it's small/fast model, but it needs to spit out a ton of tokens to work that well, which may cancel out some of the performance and memory benefits.
Something interesting: nvidia is potentially struggling to make tensor cores faster. part of why they've added 4- and 8-bit precisions is that they were unable to really make 16-bit tensor cores faster (according to someone at nvidia)
Literally 3 days before openai announced the navier-stokes stuff, (because there were rumors about it) I was commenting to someone "if someone actually uses ai to prove navier-stokes, I will consider LLMs one of the more substantial technological developments ever."
(granted they found a counter-example, which is perhaps less impressive, since counter-examples seem to be the more common success story for LLMs looking at big problems, since they are presumably easier to "brute-force." + some of the drama w.r.t. the external math researchers uploading their work into chatgpt, and it possibly being used to train the new model...)
[quote=quintosh]
some analysts predict that local AI models can be as good as AI that ran in a datacenter half a year ago, it's not just the hardware improving but several other factors too like compression techniques and algorithmic efficiencies. A GTX 3090 can run qwen 3.8-27b which gets a better AI intelligence score (doesn't just include googling information) than Claude Opus 4.5, GPT-5.2 and Gemini 3 7 months ago[/quote]
IIRC, qwen has to use a ton of tokens to get that quality on the benchmark. So it's small/fast model, but it needs to spit out a ton of tokens to work that well, which may cancel out some of the performance and memory benefits.
Something interesting: nvidia is potentially struggling to make tensor cores faster. part of why they've added 4- and 8-bit precisions is that they were unable to really make 16-bit tensor cores faster (according to someone at nvidia)
(granted they found a counter-example, which is perhaps less impressive, since counter-examples seem to be the more common success story for LLMs looking at big problems, since they are presumably easier to "brute-force." + some of the drama w.r.t. the external math researchers uploading their work into chatgpt, and it possibly being used to train the new model...)
[quote=quintosh]
some analysts predict that local AI models can be as good as AI that ran in a datacenter half a year ago, it's not just the hardware improving but several other factors too like compression techniques and algorithmic efficiencies. A GTX 3090 can run qwen 3.8-27b which gets a better AI intelligence score (doesn't just include googling information) than Claude Opus 4.5, GPT-5.2 and Gemini 3 7 months ago[/quote]
IIRC, qwen has to use a ton of tokens to get that quality on the benchmark. So it's small/fast model, but it needs to spit out a ton of tokens to work that well, which may cancel out some of the performance and memory benefits.
Something interesting: nvidia is potentially struggling to make tensor cores faster. part of why they've added 4- and 8-bit precisions is that they were unable to really make 16-bit tensor cores faster (according to someone at nvidia)
Bob_Marleynavier-stokes stuff
even if they didnt directly use the data from the other group, having that rumour anout the approach spread around would have been an insane prompting boost (human input)
quintoshA GTX 3090 can run qwen 3.8-27b
yeah, thats good, and i like running stuff locally. but in practice, it's more expensive* for a worse result. (basically I would be going out of my way to use a 6 month dumber model)
and it would be way slower!! (llms are slow enough already as-is). and it defo looks like there are some speed improvements on the way
[quote=Bob_Marley]navier-stokes stuff[/quote]
even if they didnt directly use the data from the other group, having that rumour anout the approach spread around would have been an insane prompting boost (human input)
[quote=quintosh]A GTX 3090 can run qwen 3.8-27b[/quote]
yeah, thats good, and i like running stuff locally. but in practice, it's more expensive* for a worse result. (basically I would be going out of my way to use a 6 month dumber model)
and it would be way slower!! (llms are slow enough already as-is). and it defo looks like there are some speed improvements on the way
even if they didnt directly use the data from the other group, having that rumour anout the approach spread around would have been an insane prompting boost (human input)
[quote=quintosh]A GTX 3090 can run qwen 3.8-27b[/quote]
yeah, thats good, and i like running stuff locally. but in practice, it's more expensive* for a worse result. (basically I would be going out of my way to use a 6 month dumber model)
and it would be way slower!! (llms are slow enough already as-is). and it defo looks like there are some speed improvements on the way
Please
sign in through STEAM
to post a comment.